https://github.com/goodmami/minigraph

Small Python module for small graph structures
https://github.com/goodmami/minigraph

graph graph-library unmaintained

Last synced: 10 months ago
JSON representation

Small Python module for small graph structures

• Host: GitHub
• URL: https://github.com/goodmami/minigraph
• Owner: goodmami
• License: mit
• Archived: true
• Created: 2015-05-12T10:20:35.000Z (about 11 years ago)
• Default Branch: master
• Last Pushed: 2015-05-15T09:35:57.000Z (about 11 years ago)
• Last Synced: 2025-06-11T03:17:03.801Z (about 1 year ago)
• Topics: graph, graph-library, unmaintained
• Language: Python
• Size: 145 KB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# minigraph

Small Python module for small graph structures

### Motivation

I [work with](https://github.com/goodmami/pydelphin) small graphs. I

build them programatically from other representations. I don't work with

huge graphs with millions of nodes and edges, but usually those with

less than 100. I want this in a graphing library:

  - it is fast to build graphs and make subgraphs

  - it is easy to query for basic properties, like size

  - it is fast to determine if a graph is connected or not

  - it is easy to do pathfinding

  - one graph type handles directed edges, undirected edges,

    labeled edges, and overlapping edges (multigraphs)

  - ideally it is a single file, so I can just copy it into other

    projects instead of making a dependency, which means...

  - it has a permissive open-source license

Consider some of the existing Python libraries for graphs:

  - [NetworkX][] is featureful, but slow

  - [graph-tool][] is fast, but GPL'd (it's based on the

    [Boost Graph Library][], which has a permissive license, but the

    author of graph-tool [seems uninterested in changing

    graph-tool's license](https://git.skewed.de/count0/graph-tool/issues/194))

  - [NetworKit][] is supposedly fast (though its hard to find any

    benchmarks to support this claim), and there's little documentation

    for building graphs programatically.

### Benchmarks

I compared MiniGraph to NetworkX. It's not directly comparable, because

the single MiniGraph class can be like NetworkXs Graph, DiGraph,

MultiGraph, and MultiDiGraph classes. Because of this, I compare it to

all of them. Here's the current state (times are in seconds):

    Building a chain of 100 nodes 10000 times:

      MiniGraph (regular   ) : 2.078411

      MiniGraph (fast_init ) : 1.903894

      NetworkX Graph         : 2.422691

      NetworkX DiGraph       : 3.009594

      NetworkX MultiGraph  : 4.788977

      NetworkX MultiDiGraph  : 5.369203

    Building a fully-connected 10-node graph 10000 times:

      MiniGraph (regular   ) : 1.174832

      MiniGraph (fast_init ) : 1.014850

      NetworkX Graph     : 1.749765

      NetworkX DiGraph     : 1.912611

      NetworkX MultiGraph: 4.029920

      NetworkX MultiDiGraph: 4.320380

    Building a 4x-connected 5x5-node graph 10000 times:

      MiniGraph (regular   ) : 1.068538

      MiniGraph (fast_init ) : 0.902589

      NetworkX MultiGraph: 3.878337

      NetworkX MultiDiGraph: 4.003757

A good start, but there's more work to do.

[NetworkX]: https://networkx.github.io/

[graph-tool]: https://graph-tool.skewed.de/

[Boost Graph Library]: http://www.boost.org/doc/libs/release/libs/graph

[NetworKit]: https://networkit.iti.kit.edu/